You are deep in a thread, and ChatGPT stops: "You've reached the maximum length for this conversation, but you can keep talking by starting a new chat."

Do not start a new chat yet. If you do, everything you built in that thread stops being available to the model, and ChatGPT cannot read it back. There is a better move, and it takes about ten seconds.

Quick answer: Scroll up several messages before the error, click the three dots under an earlier ChatGPT reply, and select Branch in a new chat. You get a new thread that keeps the conversation history up to that point with a fresh context budget. Do not branch from the message that errored.

First, check which limit you actually hit

ChatGPT produces two errors that people constantly confuse. They have opposite fixes, so spend five seconds identifying yours.

	Conversation length limit	Message rate limit
The message	"Maximum length for this conversation"	"You've reached our limit of messages"
What it means	This thread filled its context window	Your account sent too many messages too fast
New chat fixes it?	Yes	No
Waiting fixes it?	No, never	Yes, usually about 3 hours
Plus fixes it?	No	Often

If you are reading "maximum length," you hit the first one and this guide applies. If you are reading "limit of messages," no amount of branching or new chats will help. Switch to a lighter model or wait for the window to roll over.

Fix it: branch, don't restart

ChatGPT added branching precisely for this. It forks the conversation at a chosen point into a new thread that carries the prior history but starts fresh on budget.

1. Stay in the broken thread. Do not delete it and do not start a new chat.
2. Scroll up several messages. This is the part people get wrong. Do not branch from the reply where the error appeared, because that point is already at the ceiling. Go back a handful of exchanges to somewhere comfortably under it.
3. Click the three dots beneath that earlier ChatGPT response.
4. Choose "Branch in a new chat."
5. Rename the new thread something like "Project, part 2" so you can navigate between them later.

The new conversation opens with the context from before the branch point and room to keep working. Anything after your branch point is left behind in the original thread, which is why choosing a point that still contains your important decisions matters.

If branching is not available

Branching has not always been present on every surface and every account. If you do not see the option, fall back to a handoff summary.

Ask this in the dying thread, before you do anything else:

Create a handoff summary I can paste into a new chat. Include: what we're trying to accomplish, key decisions we've made, anything important you'd get wrong by guessing, and the very next step. Keep it short enough to read in under a minute.

That phrasing comes from PCWorld's Ben Patterson, and it beats "summarize this chat" because it asks for decisions and traps rather than a recap.

Then:

• Save it to a file, not just your clipboard. Clipboards get overwritten.
• Paste critical values verbatim. Summaries paraphrase, and paraphrasing corrupts exact numbers, names, and code.
• Open the new chat, paste the summary first, and add: "Continue from here. Confirm you understand before proceeding."

Even at the ceiling, ChatGPT can usually generate one final response, so this works as a last resort after the error appears. It just works considerably better before.

Why Plus does not fix this

This is the most expensive misconception in the topic, and most articles ranking for this error get it wrong.

A user on GenAI Stack Exchange described tracking their gym progress in a single long thread, hitting the wall, and upgrading to the paid plan on the assumption it would extend the conversation. It had no effect. They then tried opening a new chat and asking it to recall the previous thread, and got told ChatGPT has no access to it, followed by workout suggestions that ignored all their prior progress.

The distinction is that paid plans raise rate caps, how many messages you may send in a window. The per-conversation context ceiling is a property of the model, and it applies the same way whether you pay or not.

One more caution worth knowing: a June 2026 report on OpenAI's community forum describes threads that became too long turning unreadable, redirecting to the homepage with "The conversation is too long, please start a new one." Not merely uncontinuable, inaccessible. Treat a long thread as something to extract from early rather than an archive you can rely on.

Why does ChatGPT forget mid-thread, before any error?

Sometimes there is no error, just decay. You set a format rule at the start, and forty messages later it is being ignored.

That is context rot. The model weights recent tokens more heavily than distant ones, so early instructions are technically present but assigned low importance. It is reading them and effectively skipping them.

Three things help:

• Re-anchor every ten or so exchanges. Paste a short recap of the decisions that still apply.
• Put critical constraints at the end of your message, closest to the question, where they carry the most weight.
• Split unrelated tasks into separate chats. Mixing a coding task and a writing task in one thread makes the model balance two contexts at once.

Note that Memory does not solve this either. Memory stores durable facts about you, your name, your preferences, your ongoing projects. It does not store the plot of your novel or the state of your refactor.

Stop having this problem

Everything above is you doing ChatGPT's filing: notice the thread is long, remember to ask for a summary, store it, re-paste it, hope nothing important got paraphrased away. Miss the moment and it is gone.

The root cause is architectural. Your context lives inside a thread, inside one app, on one device. The thread is the unit of memory, so when it ends the memory ends with it.

folk is built the other way around. It compacts long conversations automatically, and more importantly its memory is not scoped to a thread at all. It reaches you in iMessage and Telegram, so there is no conversation to reach the end of and nothing to re-paste when one does. What you told it last month is still there.

If you have context in ChatGPT worth keeping, you do not need to retype it. You can import your ChatGPT and Claude memory in a single paste.

Hitting this on a different assistant? See the guides for Claude and Gemini, or the overview of why every AI assistant forgets.